INDEX

Explanations

phrases beginning with informational headers often used in articles or online content

oai_token-act-pair · gpt-3.5-turbo

instances of introductory phrases or sentences in written content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

iously

-0.66

ishly

-0.65

 Others

-0.64

isSpecialOrderable

-0.64

inately

-0.62

tsky

-0.61

ãħĭãħĭ

-0.60

afi

-0.59

Others

-0.59

osh

-0.59

POSITIVE LOGITS

 however

0.96

 meanwhile

0.75

 wherein

0.73

 moreover

0.72

 please

0.62

 Kinnikuman

0.61

 respectively

0.59

 though

0.59

 albeit

0.59

where

0.59

Activations Density 0.423%

phrases beginning with informational headers often used in articles or online content

instances of introductory phrases or sentences in written content

No Comments

No Known Activations